a) Antibody Crystals
With over 100 monoclonal antibodies (mAbs) currently being evaluated in clinical study phases 2 or 3, the mAb market is considered one of the most promising biopharmaceutical markets. Since these drugs are delivered in single doses often exceeding 100 mg, there is an urgent need to find suitable formulation strategies that satisfy stability, safety, and patient compliance. However, highly concentrated liquid mAb formulations show increased viscosity, hindering syringeability through patient friendly thin needles. Furthermore, the tendency for mAb molecules to aggregate at such high concentrations exponentially increases when compared to moderately concentrated solutions. This is unacceptable, with regard to safety and stability requirements.
Thus, the delivery of high mAb doses is reserved for large volumes, which generally have to be delivered via infusion. This way of dosing is cost intensive and significantly reduces the patient's compliance.
Therefore, pharmaceutically applicable low volume mAb crystal suspensions for subcutaneous injection would be highly desirable. Theoretically, degradation pathways influencing the mAb integrity should be significantly decelerated due to the rigidity of a crystal lattice, where motions in the protein structure are hindered. Moreover, an increase in viscosity would be significantly reduced when comparing highly concentrated crystal suspensions with liquid formulations. With respect to sustained release, it might be possible to generate or alter protein crystals such that they dissolve slowly when brought into a patient's body. This would be a very elegant way to deliver a sustained release formulation, as the extensive use of excipients and processes harming the mAb structure would be prevented.
Despite the great potential in using protein crystals as a drug substance, few attempts have been made to systematically evaluate this strategy.
A well-known exemption is insulin, which was successfully crystallized decades ago. Today, the use of crystal suspensions of insulin is well described, offering stable and long acting formulations being well established on the market. The discrepancy between the development of insulin crystals and crystallization of all other proteins might be related to the fact that ordered insulin aggregates are natively formed in the pancreas. Thus, insulin crystals are easily obtained when insulin is brought in contact with an excess of zinc ions. Most other proteins tend to form unordered precipitates rather than crystals, and therefore, finding crystallization conditions for a protein is a time consuming, non-trivial task.
Despite a great interest in harvesting protein crystals for x-ray diffraction analysis, finding suitable crystallization conditions still is an empirical science, as in principle any protein behaves differently. To date, no general rule has been found that might reliably predict a successful crystallization condition for a protein of choice. Thus, obtaining crystals of a given protein always is referred to as the “bottle neck” of whatever intended application is planned later on.
Antibodies are especially hard to crystallize, due to the flexibility of the molecule. Nevertheless, examples of immunoglobulin crystals have been known for a long time. The first example of immunoglobulin crystals were described 150 years ago by an English physician, Henry Bence Jones; he isolated crystals of an abnormal Ig light chain dimer from the urine of a myeloma patient (Jones, H. B. (1848) Philosophical Transactions of the Royal Society, London 138: 55-62). Such abnormal Igs have been known ever since as Bence Jones proteins. In 1938, the spontaneous crystallization of a distinct abnormal Ig from the serum of a myeloma patient was described (von Bonsdorf, B. et al. (1938) Folia Haematologia 59: 184-208), apparently an Ig heavy chain oligomer (MW 200 kDa).
Crystalline human immunoglobulins of normal structure (two heavy chains linked to two light chains) were described over the next thirty years, again mostly isolated from myeloma patients (Putnam, F. W. (1955) Science 122: 275-7). Davies and co-workers were the first to characterize the structure of an intact human myeloma antibody, named “Dob”, using x-ray crystallography (Terry, W. D. et al. (1968) Nature 220(164): 239-41), and they determined its three-dimensional structure in 1971 (Sarma, V. R. et al. (1971) J. Biol. Chem. 246(11): 3753-9). Their pioneering work was followed by that of others, yielding the crystal structures of the IgG “Kol” (Huber, R. et al. (1976) Nature 264(5585): 415-20), the IgG “Mcg” (Rajan, S. S. et al. (1983) Mol. Immunol. 20(7): 787-99), and a canine lymphoma IgG2a (Harris, L. J. et al. (1992 Nature 360(6402): 369-72).
Crystals of immunoglobulins retain their distinctive immunological activities upon re-dissolution. Nisonoff et al. reported in 1968 on a rabbit anti-p-azobenzoate antibody, “X4”, that was easily crystallized (Nisonoff, A. et al. (1968) Cold Spring Harbor Symposia on Quantitative Biology 32: 89-93). Antibody X4 was extensively characterized before crystallization as well as after re-dissolution of the crystals. [125I]-p-iodobenzoate was found to bind specifically and potently to re-dissolved X4; the re-dissolved crystals also exhibited multiple specific Ouchterlony immunodiffusion reactions typical of the unpurified rabbit serum (Nisonoff et al., 1968). Connell and co-workers described a human myeloma gamma-immunoglobulin-1 kappa (IgG-κ), called “Tem”, that crystallized spontaneously from serum at cold temperatures (Connell, G. E. et al. (1973) Canad. J. Biochem. 51(8): 1137-41). Tem crystals were found to be well-formed and possessed rhombohedral symmetry. Tem-containing serum was extensively characterized by agarose immunodiffusion techniques. Electrophoresis and immunodiffusion of a re-dissolved solution of the Tem crystals showed them to be identical with the material obtained from the serum by cryoprecipitation, and with the isolated myeloma protein (Connell et al., 1973).
Mills and co-workers reported in 1983 an unusual crystallocryoglobulinemia resulting from human monoclonal antibodies to albumin (Mills, L. E. et al. (1983) Annals of Internal Med 99(5): 601-4). Here, very similar cuboidal crystals were isolated from two patients. Redissolution of the crystals followed by electrophoresis and immunoelectrophoresis indicated that the crystals were composed of two protein components, a monoclonal IgG-lambda and human serum albumin in a 1:2 ratio (Jentoft, J. E. et al. (1982) Biochem. 21(2): 289-294). The components were separated on preparative scale by dissolution of the original crystals followed by column chromatography. Although neither separated component crystallized on its own, upon recombination the original bipartite complex reformed and then crystallized. Further study of the distinctive sedimentation characteristics and immunological reactivity of the redissolved, separated IgG and its Fab fragment with human serum albumin indicated that reassociation of the two redissolved, separated components was immunologic in nature, i.e., that the crystalline antibody once redissolved still possessed its native, highly specific (for human serum albumin) binding characteristics (Mills et al. 1983).
Recently, Margolin and co-workers reported on the potential therapeutic uses of crystalline antibodies (Yang, M. X. et al. (2003) Proc. Natl. Acad. Sci. 100(12): 6934-6939). They found that the therapeutic monoclonal antibody trastuzumab (Herceptin®) could be crystallized (Shenoy, B. et al. (2002) PCT Int. Appl. WO/2002/072636, (Altus Biologics Inc., USA). 173 pp.). Crystalline trastuzumab suspensions were therapeutically efficacious in a mouse tumor model, thus demonstrating retention of biological activity by crystalline trastuzumab (Yang et al., 2003).
b) Crystallization Techniques
The crystallization of diverse proteins cannot be carried out successfully using defined methods or algorithms. Certainly, there have been great technical advances in the last 20-30 years, as noted by the world-renowned expert in protein crystallization, A. McPherson. McPherson provides extensive details on tactics, strategies, reagents, and devices for the crystallization of macromolecules. (McPherson, A. (1999) Crystallization of Biological Macromolecules. Cold Spring Harbor, N.Y., Cold Spring Harbor Laboratory Press, p. 159). He does not, however, provide a method to ensure that any given macromolecule can indeed be crystallized by a skilled person with reasonable expectation of success. McPherson states for example: “Whatever the procedure, no effort must be spared in refining and optimizing the parameters of the system, both solvent and solute, to encourage and promote specific bonding interactions between molecules and to stabilize them once they have formed. This latter aspect of the problem generally depends on the specific chemical and physical properties of the particular protein or nucleic acid being crystallized.”.
It is widely accepted by those skilled in the art of protein crystallization that no algorithm exists to take a new protein of interest, apply definite process steps, and thereby obtain the desired crystals.
Several screening systems are commercially available (for example Hampton 1 and 2, Wizzard I and II) which allow, on a microliter scale, to screen for potentially suitable crystallization conditions for a specific protein. However, positive results obtained in such a screening system do not necessarily allow successful crystallization in a larger, industrially applicable batch scale. Conversion of microliter-size crystallization trials into industrial dimensions is described to be a challenging task (see Jen, A., Merkle, H. P. (2001) Pharm. Res. 18, 11, 1483).
Baldock et al. reported on a comparison of microbatch and vapor diffusion for initial screening of crystallization conditions (Baldock, P. et al. (1996) J. Crystal Growth 168(1-4):170-174. Six commercially available proteins were screened using a set of crystallization solutions. The screens were performed using the most common vapor diffusion method and three variants of a microbatch crystallization method, including a novel evaporation technique. Out of 58 crystallization conditions identified, 43 (74%) were identified by microbatch, while 41 (71%) were identified by vapor diffusion. Twenty-six conditions were found by both methods, and 17 (29%) would have been missed if microbatch had not been used at all. This shows that the vapor diffusion technique, which is most commonly used in initial crystallization screens does not guarantee positive results.
c) Anti-Human IL-12 Antibody Crystals
Human IL-12 plays a critical role in the pathology associated with several diseases involving immune and inflammatory responses, for example multiple sclerosis, Crohn's disease and psoriasis. There is, therefore, a great need for suitable methods of treating such human IL-12 related disorders. One promising therapeutic approach comprises the administration of pharmaceutically effective doses of anti-human IL-12 antibodies.
Due to the role of human IL-12 in a variety of human disorders, therapeutic strategies have been designed to inhibit or counteract IL-12 activity. In particular, antibodies that bind to, and neutralize, IL-12 have been sought as a means to inhibit IL-12 activity. Some of the earliest antibodies were murine monoclonal antibodies (mAbs), secreted by hybridomas prepared from lymphocytes of mice immunized with IL-12 (see, e.g., WO 97/15327). These murine IL-12 antibodies are, however, limited for their use in vivo due to problems associated with administration of mouse antibodies to humans, such as short serum half life, an inability to trigger certain human effector functions and elicitation of an unwanted immune response against the mouse antibody in a human (the “human anti-mouse antibody” (HAMA) reaction).
In general, attempts to overcome the problems associated with the use of fully-murine antibodies in humans, have involved genetically engineering the antibodies to be more “human-like.” For example, chimeric antibodies, in which the variable regions of the antibody chains are murine-derived and the constant regions of the antibody chains are human-derived, have been prepared. However, because these chimeric and humanized antibodies still retain some murine sequences, they still may elicit an unwanted immune reaction, the human anti-chimeric antibody (HACA) reaction, especially when administered for prolonged periods.
U.S. Pat. No. 6,914,128 discloses human antibodies, preferably recombinant human antibodies, that specifically bind to human interleukin-12 (hIL-12). Preferred antibodies disclosed therein, have high affinity for hIL-12 and neutralize hIL-12 activity in vitro and in vivo. The antibodies, or antibody portions, are useful for detecting hIL-12 and for inhibiting hIL-12 activity, e.g., in a human subject suffering from a disorder in which hIL-12 activity is detrimental. Nucleic acids, vectors and host cells for expressing the recombinant human antibodies of the invention, and methods of synthesizing the recombinant human antibodies, are also enclosed. Crystalline forms of the anti-hIL-12 antibodies or methods for preparing the same are not specifically described in the '128 patent.
The problem to be solved according to the present invention is, therefore, to develop suitable crystallization conditions, in particular batch crystallization conditions, for anti-IL-12 antibodies, and to establish crystallization process conditions applicable to volumes relevant for industrial antibody crystal production. At the same time, a crystallization process is established that does not make use of toxic agents, which might negatively affect the pharmaceutical applicability of such antibodies.